Roland Juno-106, MKS-30, and GR-700
Failure mechanism: The Roland 800017A VCF/VCA hybrid IC suffers from internal metal migration, which eventually causes internal shorts.
Failure signature: A voice crackles, cuts in and out, or continues to sound when no key is pressed. Rustling noises may be heard when playing or when no key is pressed. Less commonly, a filter may behave erratically or refuse to stay in calibration. Symptoms may get either better or worse as the synth warms up. On the Juno-106 or the MKS-30, in Poly 1 mode, every sixth note played exhibits the symptoms. On the GR-700 guitar synth, one particular string exhibits the symptoms.
Advice for buyers: If you can get access to the synth, power it up in test mode (or tuning mode for the GR-700)and test all six voices. Do this once when the synth is cold, and again after it has warmed up for about ten minutes. If you can't get access to the synth (e.g., an Ebay deal), insist that the seller disclose whether or not all six voices work and stand behind his word.
Advice for owners: Despite some Internet speculation, no one has managed to identify any particular date codes that will or won't fail. Owners are advised to stockpile a couple of spares. Currently 80017A's are going for about $65 US on Ebay. An alternative is to buy a broken 106 for parts; a used 106 with problems will often sell for less than the cost of a full set of 80017A's. Another alternative is a clone part. There are several available, but the one I recommend is the D80017 from Analog Renaissance. It's nearly the same size as the real thing, so it won't cause any clearance or heating problems inside the case, and the manufacturer claims that calibration trimming will be similar.
Difficulty of repair: Not difficult. Anyone with soldering skills should be able to do it. The board is clearly labeled as to which ICs go with which voices. For the Juno-106, see my previous post for the location and layout of the module board. For the MKS-30 and the GR-700 (which use the same board), see this page from Wayne Scott Jones.
Failure mechanism: A NiCad memory backup battery installed at the factory leaks onto the CPU board, eating the copper traces off the board and damaging the substrate.
Failure signature: Random behavior; panel buttons don't work, lights flash randomly, voices don't play. Or, the synth doesn't boot up at all. Note that serious damage can be done before the battery fails to the point where patch memory doesn't retain.
Advice for buyers: Ask the seller if the battery has been replaced and the CPU board inspected and/or repaired. If the answer is no, or if the seller doesn't know what you are talking about, don't buy unless you have a good tech or advanced electronics repair skills.
Advice for owners: If you don't know that your Polysix has been previously repaired and a lithium battery fitted, open it up and inspect the CPU board. This page from "The Old Crow" (Scott Rider) shows how to open the synth up and what to look for; it also contains the detailed repair procedure.
Difficulty of repair: Replacing the battery, in itself, is not difficult. However, by now, chances are that any Polysix that still has its original battery has suffered significant damage to the CPU board. The battery acid can be neutralized by cleaning the board with baking soda and water, but damaged board traces have to be identified and jumpered. It's a tedious job that requires some knowledge of the circuits, and advanced soldering skills.
Update: Commenter "steamy vicks" advises that the Kawai SX-240, and possibly also the SX-210, suffers from this same issue.
Rhodes Chroma Polaris
Failure mechanism: Ribbon cables which are molded into the membrane panel have been embrittled with age, and crack or split when moved or subject to vibration.
Failure signature: Some or all panel buttons don't work.
Advice for buyers: Frankly, unless you absolutely have to have one, or you are handy with electronics and you want something to experiment on, don't buy. If you do find one cheap, be aware that the Curtis IC's in the synth are probably worth more than the synth itself.
Advice for owners: Don't gig with your Polaris; in fact, don't move it at all if you can avoid it. Don't open up the synth unless you have to; opening it flexes the ribbon cables.
Difficulty of repair: Because the ribbon cables are one piece with the membrane panel, the easiest path of repair is to replace the whole membrane. This is not a difficult repair; the difficult part is finding a replacement. Here is a list of service centers from rhodeschroma.com, but it is my understanding that there aren't many Polaris panels left. And, be aware that any replacement panels that do exist are as old as the originals, and are likely to also have the embrittlement problem. Attempts to splice the ribbons back together usually fail because handling the ribbons cause them to crack further. In this thread from Harmony Central, a poster succeeded in soldering wires onto the end of the ribbon cable and connecting them to a bulkhead connector. However, as you can see from reading the thread, it was a very exacting job requiring, among other things, a temperature-controlled soldering iron. The best way to address the problem appears to be to just use a software editor and edit via MIDI. I've seen talk about designing a replacement panel using real switches, but I haven't found anyone who has actually done this.
Failure mechanism: The LM2940 voltage regulator on the DSP board shorts out and applies a high voltage to the DSP board, which fries the processors on the board.
Failure signature: The synth displays an "ESP" message in the display on power up, and won't play.
Advice for buyers: Ask the seller if the LM2940 has been replaced. If the answer is no, or if the seller doesn't know, do the following: (1) Make the seller assure you that the synth was working the last time it was powered up; (2) Instruct the seller to not power it up again before shipping, and (3) when you receive it, do not power it up until you inspect it and replace the regulator if necessary.
Advice for owners: If you don't know for sure that the regulator in your Fizmo has been replaced, do not power it up again until you have inspected it and replaced the regulator if necessary. If you do know that the regulator has been replaced, you're home free. Note that apparently some Fizmos had th regulator replaced at the factory before shipping; if the regulator in your Fizmo is something other than an LM2940, it's probably OK. Also note: There are a number of threads in various places speculating as to whether the factory supplied wall wart was responsible for these failures. This turned out to be not true, but there are Fizmos around with wall warts that don't supply enough current, which can cause erratic behavior. See Carbon111's page referenced below for more information.
Difficulty of repair: It depends on whether or not the LM2940 has failed. If the synth is still working, anyone who is handy with a soldering iron and a drill can replace the regulator. Carbon111 has an excellent page showing how to inspect the regulator, and it contains a link to a page with the detailed repair procedure. Unfortunately, if the regulator has failed and the synth is displaying the "DSP" error, repair is all but impossible. The DSP's are small-pitch surface mount parts that cannot be soldered or unsoldered without specialized equipment and advanced skills, and the cost of the repair would likely exceed the synth's market value. There is no known source for replacement DSP boards.